The White House is considering providing U.S. federal agencies access to a version of Anthropic’s advanced AI model, Mythos, as part of efforts to strengthen cybersecurity, Bloomberg reported.
Announced earlier this month, Mythos is being tested under Anthropic’s “Project Glasswing,” a controlled initiative that allows select organizations to use a preview version of the model for defensive cybersecurity purposes.
The tool has reportedly identified thousands of vulnerabilities across operating systems, web browsers, and other software, highlighting both its potential and associated risks.
According to Bloomberg, Gregory Barbaccia of the Office of Management and Budget (OMB) informed federal officials that safeguards are being developed to enable agency access.
“We’re working closely with model providers, other industry partners, and the intelligence community to ensure the appropriate guardrails and safeguards are in place,” he said.
While no timeline or formal rollout has been confirmed, officials emphasized the need for careful evaluation before deployment. The initiative reflects broader government efforts to partner with AI developers to secure critical systems.
The discussions come amid ongoing engagement between Anthropic and the administration, even as the Department of Defense recently ended a contract with the company following a dispute